feat(AggLayer claim e2e #4): e2e bridge-in flow with real bridge data by mmagician · Pull Request #2413 · 0xMiden/protocol

mmagician · 2026-02-08T21:04:22Z

Changes:

unify loading of all ProofData elements (exit roots, merkle paths) to use LE ordering
unify procedure locals to also use LE ordering (this is mostly for consistency with the above; it doesn't matter if we do loc_storew_be ... loc_loadw_be or loc_storew_le ... loc_loadw_le)
use "real" bridge-in data
- I made a transfer from L1 -> Katana which triggered a claimAsset tx on Katana on Katana to a Miden-like address 0x00000000b0E79c68cafC54802726C6F102Cca300
- this would be very similar to a claim on Miden; the only difference would be the chain ID
- for compatibility testing, the address on L1 must be decodable into a Miden AccountId

This concludes the PR tower: we now fully verify the data inside the CLAIM note.
What's missing for a full integration is actually checking that the GER we compute is part of onchain storage, which is tackled separately in #2388

mmagician · 2026-02-08T22:00:30Z

crates/miden-agglayer/solidity-compat/test/ClaimAssetTestVectors.t.sol

a lot of the test vectors have duplicated data, maybe we could eventually create one big test vector, but it's not a priority right now

crates/miden-testing/tests/agglayer/crypto_utils.rs

Fumuran

Looks great! I left a few questions and proposals inline.

crates/miden-agglayer/asm/bridge/crypto_utils.masm

Fumuran · 2026-02-10T15:31:08Z

crates/miden-agglayer/asm/bridge/crypto_utils.masm

    # Merkle path is guaranteed to contain 32 nodes
    repeat.32
-        # load the Merkle path node onto the stack
-        mem_stream


I wonder do we have an incorrect documentation for the mem_stream instruction or it was just already updated to work with the latest VM which uses LE.

If we have [1, 2, 3, 4, 5, 6, 7, 8] memory layout, where 1 is placed at memory address 0 and 8 at memory address 7, then the mem_stream will load this double word onto the stack in the reversed order: [8, 7, 6, 5, 4, 3, 2, 1], where 8 is placed at the top. At the same time in the mem_stream docs it is written:

Stack transition: [C, B, A, a, ... ] -> [E, D, A, a', ... ], where
[E, D] ← [mem[a..(a+4)], mem[(a+4)..(a+8)]], a' ← a+8.

So it describes the transition in LE, not BE.

Also the necessity of whiting the LE mem_stream by your own made me think whether should we create the LE version of the mem_stream (the same way it was done with mem_storew_le and the others) and rename the current mem_stream into the mem_stream_be.

crates/miden-agglayer/asm/bridge/crypto_utils.masm

crates/miden-testing/tests/agglayer/test_utils.rs

partylikeits1983

This is awesome! Looks great!

Great to see the actual claim in bridge data being finally used in the claim note test!!!!

partylikeits1983 · 2026-02-13T17:25:00Z

crates/miden-testing/tests/agglayer/bridge_in.rs

+    // Note: We intentionally do NOT verify the exact note ID or asset amount here because
+    // the scale_u256_to_native_amount function is currently a TODO stub that doesn't perform
+    // proper u256-to-native scaling. The test verifies that the bridge-in flow correctly
+    // validates the Merkle proof using real cryptographic proof data and creates an output note.
+    //
+    // TODO: Once scale_u256_to_native_amount is properly implemented, add:
+    // - Verification that the minted amount matches the expected scaled value
+    // - Full note ID comparison with the expected P2ID note
+    // - Asset content verification


Just FYI, the procedure that checks the u256 amount to Felt scaling amount needs the Felt amount to be provided via NoteStorage

crates/miden-testing/tests/agglayer/test_utils.rs

crates/miden-testing/tests/agglayer/bridge_in.rs

chore: add claim test gen to makefile

… compatibility The keccak256::merge MASM function applies word::reverse to its input words before hashing. mem_stream (used to load SMT proof nodes) does NOT apply per-word reversal, while loc_loadw_be (used to load the current hash) DOES. To produce consistent per-word reversed format on the stack for both the current hash (via loc_storew_be/loc_loadw_be) and the path nodes (via mem_stream), SMT proof nodes must be stored in fully reversed element order in memory. Exit roots are per-word reversed because they are loaded via mem_load_double_word (which uses mem_loadw_be), and the per-word reversal by mem_loadw_be undoes the storage reversal to produce natural order on the stack. Also adds diagnostic and integration tests for Merkle proof verification using claim_asset_vectors data. Co-authored-by: marti <marti@hungrycats.studio> fix: simplify bridge_in test assertions for stubbed scale function The test now correctly verifies that the bridge-in flow validates the Merkle proof and creates an output note, without asserting specific note content that depends on the stubbed scale_u256_to_native_amount function. Removes diagnostic test test_claim_asset_verify_leaf_via_advice_map which has a stack management issue unrelated to the core fix. Co-authored-by: marti <marti@hungrycats.studio> chore: fix warnings and remove diagnostic test - Remove unused imports (Asset, NoteRecipient, NoteStorage) - Remove unused leaf_elements variable - Remove test_claim_asset_verify_leaf_via_advice_map diagnostic test (covered by test_bridge_in_claim_to_p2id) Co-authored-by: marti <marti@hungrycats.studio> refactor: use data-level fix without MASM changes The keccak256::merge function (from external miden-core-lib) applies word::reverse internally and expects per-word reversed input. Since loc_loadw_be naturally produces per-word reversal, changing to loc_loadw_le would break merge's byte ordering. Instead, fix at the data level: - SMT proof nodes: fully reversed (to_memory_elements) so mem_stream's half-swap produces per-word reversed format matching loc_loadw_be output - Exit roots: per-word reversed (to_word_reversed_elements) so mem_load_double_word's mem_loadw_be reversal produces natural order - No MASM code changes needed Co-authored-by: marti <marti@hungrycats.studio> chore: remove diagnostic tests Remove 7 diagnostic/exploratory tests that were added during debugging. The pre-existing tests (pack_leaf_data, get_leaf_value, test_solidity_verify_merkle_proof_compatibility) plus the fixed test_bridge_in_claim_to_p2id provide sufficient coverage. Co-authored-by: marti <marti@hungrycats.studio>

Replace mem_stream (which requires RPO hasher state padding) with two explicit mem_loadw_be calls for loading Merkle path nodes. This uses the same instruction as loc_loadw_be for the current hash, ensuring both undergo identical element ordering. The change eliminates 12 elements of RPO padding from the stack, simplifies pointer management, and makes the memory access pattern more explicit. Co-authored-by: marti <marti@hungrycats.studio>

- Replace loc_storew_be/loc_loadw_be with loc_storew_le/loc_loadw_le - Replace mem_load_double_word with inline mem_loadw_le - Replace mem_store_double_word with inline mem_storew_le - Update canonical_zeros.masm and build.rs to use mem_storew_le Co-authored-by: marti <marti@hungrycats.studio> chore: update mem_load/store_double_word to LE and resolve TODO - Update utils.masm: mem_storew_be -> mem_storew_le, mem_loadw_be -> mem_loadw_le - Replace inline mem_loadw_le in crypto_utils.masm with exec.utils::mem_load_double_word - Restore mem_load/store_double_word usage in mmr_frontier32_keccak.masm and canonical_zeros.masm - Update stale comment in claim_note.rs Co-authored-by: marti <marti@hungrycats.studio>

mmagician force-pushed the mmagician-global-index-wrapper branch from 19cb775 to 778b6e7 Compare February 8, 2026 21:33

mmagician changed the title ~~feat(AggLayer): e2e bridge-in flow with real bridge data~~ feat(AggLayer claim e2e #4): e2e bridge-in flow with real bridge data Feb 8, 2026

mmagician force-pushed the mmagician-real-bridge-data branch from be37f12 to 43fbf27 Compare February 8, 2026 21:39

mmagician added the agglayer PRs or issues related to AggLayer bridging integration label Feb 8, 2026

mmagician linked an issue Feb 8, 2026 that may be closed by this pull request

AggLayer: Use real bridge data for testing #2397

Closed

mmagician marked this pull request as ready for review February 8, 2026 21:43

mmagician mentioned this pull request Feb 8, 2026

feat(AggLayer): Store hash(GER) in the bridge storage and implement assert_valid_ger #2388

Merged

mmagician added the no changelog This PR does not require an entry in the `CHANGELOG.md` file label Feb 8, 2026

mmagician commented Feb 8, 2026

View reviewed changes

mmagician requested review from bobbinth and partylikeits1983 and removed request for partylikeits1983 February 8, 2026 22:02

mmagician linked an issue Feb 9, 2026 that may be closed by this pull request

Bridging in: AggLayerFungibleFaucet contract (only bridging in functionality) #1900

Closed

mmagician added the pr-from-maintainers PRs that come from internal contributors or integration partners. They should be given priority label Feb 9, 2026

mmagician requested a review from Fumuran February 10, 2026 11:33

Fumuran reviewed Feb 10, 2026

View reviewed changes

This was referenced Feb 10, 2026

feat(AggLayer bridge-out e2e #1): process bridging-out data and load LeafData to memory #2425

Merged

feat: u256 to felt scaling procedure #2331

Merged

partylikeits1983 approved these changes Feb 13, 2026

View reviewed changes

mmagician force-pushed the mmagician-global-index-wrapper branch from 12c5dda to df775f2 Compare February 14, 2026 17:44

Base automatically changed from mmagician-global-index-wrapper to agglayer-new February 14, 2026 18:12

mmagician and others added 8 commits February 14, 2026 18:17

feat: infra for reading full real claim data

ff51513

chore: add claim test gen to makefile

chore: move deser structs to test_utils.rs

41a91c7

feat: use real claim data in bridge in test

f24574a

chore: mem_loadw_le for merkle paths

c9d3211

chore: use LE for local word storage

cfa4c9a

chore: mem_loadw_le for expected root

f00766c

mmagician and others added 9 commits February 14, 2026 18:17

test: prepare leaf in LE-felt order

c8e2a7b

chore: simplify inline comments

1daf2d1

lints

3d7eca7

chore: re-enable bridge-in tests

e1fc980

chore: remove dead code

5eab793

remove keccak_digest_to_word_strings; use SmtRoot/ExitRoot

903ab2c

chore: avoid loc mem, keep is_odd on stack

7d852f2

lints

e8c4306

mmagician force-pushed the mmagician-real-bridge-data branch from 4bf984c to e8c4306 Compare February 14, 2026 18:25

remove unused import

bdb93c2

mmagician merged commit c5cc3d2 into agglayer-new Feb 15, 2026
15 checks passed

mmagician deleted the mmagician-real-bridge-data branch February 15, 2026 10:46

mmagician mentioned this pull request Feb 15, 2026

AggLayer: Use real bridge data for testing #2397

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(AggLayer claim e2e #4): e2e bridge-in flow with real bridge data#2413

feat(AggLayer claim e2e #4): e2e bridge-in flow with real bridge data#2413
mmagician merged 18 commits intoagglayer-newfrom
mmagician-real-bridge-data

mmagician commented Feb 8, 2026 •

edited

Loading

Uh oh!

mmagician Feb 8, 2026

Uh oh!

Uh oh!

Fumuran left a comment

Uh oh!

Uh oh!

Fumuran Feb 10, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

partylikeits1983 left a comment

Uh oh!

partylikeits1983 Feb 13, 2026

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

Conversation

mmagician commented Feb 8, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

mmagician Feb 8, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Fumuran left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Fumuran Feb 10, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

partylikeits1983 left a comment

Choose a reason for hiding this comment

Uh oh!

partylikeits1983 Feb 13, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

mmagician commented Feb 8, 2026 •

edited

Loading